Add use_memoryview param to NdrStream() for speed. Default=False as it currently require special handling in response.

windows/rpc/client.py

@@ -225,6 +225,11 @@ class RPCClient(object):
             viewattr = response.view_attribute
             assert viewattr.ViewSize >= rpcdatasize
             data = windows.current_process.read_memory(viewattr.ViewBase, rpcdatasize)
+            # We have copied the data from the view: free it
+            # It looks like:
+            # windows.winproxy.NtAlpcDeleteSectionView(self.alpc_client.handle, 0, viewattr.ViewBase)
+            # Will ne be enough, as it will wait for the message to be reused to unmap the section..
+            # And the current architecture encourage creating new message every time
         if response_header.request_id == self.REQUEST_IDENTIFIER_ORPC:
             # Parse & remove ORPC headers (orpcthat + LocalThat)
             data = self._pass_orpc_response_structures(data)

windows/rpc/ndr.py

@@ -173,6 +173,8 @@ class NdrWString(object):
         assert size1 == size2
         assert zero == 0
         s = stream.read(size1 * 2)
+        if stream.use_memoryview:
+            return bytearray(s).decode("utf-16-le")
         return s.decode("utf-16-le")
 
     @classmethod
@@ -551,10 +553,14 @@ class NdrGuidConformantArray(NdrConformantArray):
 
 class NdrStream(object):
     """A stream of bytes used for NDR unpacking"""
-    def __init__(self, data):
+    def __init__(self, data, use_memoryview=False):
+        """"""
         self.fulldata = data
-        self.data = data
-        # self.data = memoryview(data) # FAST but need some code rewrite at some places
+        self.use_memoryview = use_memoryview
+        if use_memoryview:
+            self.data = memoryview(data) # FAST but need some code rewrite at some places
+        else:
+            self.data = data
 
     def partial_unpack(self, format):
         size = struct.calcsize(format)